Recent advances in the laboratory sciences have led to the discovery of a large number of candidate biomarkers, which hold great potential for disease diagnosis and treatment. At this time, an important research bottleneck is the lack of well-developed statistical methods for effectively using these candidate biomarkers to enhance clinical practice. It is our goal to develop new tools to select, combine, and evaluate biomarkers for disease classification and treatment selection. Classification markers predict an individual's disease outcome and are useful for the detection of diseases at an early stage when a treatment is most effective. Research proposed in Aim 1 seeks to select and combine markers to improve the classification performance in disease screening and diagnosis. Treatment selection markers predict a patient's response to different therapies and allow for the selection of a therapy that has the best predicted outcome. Aim 2 seeks to develop marker-based treatment selection rules to maximize the benefit to the patient population. A biomarker that is useful for guiding treatment decision to the general population will have different values to different patients due to individual differences in their response to treatment and in their tolerance of the disease harm and treatment cost. Aim 3 seeks to develop a new graphical tool to customize the evaluation of a biomarker for aiding treatment decision based on personal characteristics. Our statistical methods will apply broadly to general medical fields. In particular, we will apply these methods to analyze several cancer studies including (1) biomarker studies for prostate cancer and pancreatic cancer from the Early Detection and Research Network; (2) the Women's Health Initiative breast cancer genome-wide association study; and (3) the Oncotype-Dx breast cancer study from the Southwest Oncology Group. Programs and algorithms developed in this proposal will be made available to public.